QCWDRSSTC
Overview
During the planing and construction of this coil I went through many revisions of individual parts and components. You can see all my design reconsiderations in the Flickr dump. The original secondary and topload still stands, as well as the goal of ideally ~1.5m arcs. I was really impressed by the arcs produced from Gao Guangyan's first 1.5 revision QCW. The arcs produced from a secondary with a resonance frequency of ~300kHz are in my opinion, more visually appealing than coils of a higher resonance frequency. Although I might be sacrificing some arc length by using this design I think it's worth it for the more fractal but still sword-like sparks.
Specifications:
Dual full bridge of TO-247 IGBTs
6" secondary with 34 AWG wire and a resonant freq with topload of 390kHz (upper pole) 340kHz-360kHz simulated streamer.
Primary coil undetermined. Still need to tune and stuff.
Topload is 250mm x 60mm spun aluminum toroid
10000uF bus capacitance
Buck Modulator
I've noticed that a lot of people tend to glance over this part of the QCW. While a design like this is not crucial for operation, it does help to reliably control the buck converter and in turn the bus of the inverter bridge. I'd like to take some time to go over the design and operation of my buck modulator. The modulator works on a principle called "delta modulation".
Power: It's important to consider the power source you'll be using for the modulator. Since the gate driver IC will be driving the gate of the buck converters switch it will be exposed to the full bus voltage of a few hundred volts with no galvanic isolation. The FOD3184 was chosen because its internal construction features optical isolation from the input side to the output. The chip still requires 15V of power that will be non-isolated from the high voltages experienced on the output. This means there will have to be 2 isolated sources of power. One for powering the logic, and another for the output. In this design this is achieved by utilizing a isolated DC-DC converter for powering the logic. This means that we only need to use one power input for powering the entire modulator as opposed to some other designs that use 2 separate power supplies.
Logic: The start of the circuit is the IF-D95T fiber optic receiver. The output from the receiver is then passed through a Schmitt trigger to clean it up. The next section is used to fine tune and define the ramp. This process involves passing it through an op amp to integrate it and then into a comparator to convert it into the signal that will feed into the gate drive IC.
Over Current Detection: My approach to OCD is instead of using a more discrete option like a LM311 and external latch, I use MAX835 latching voltage monitor for a completely integrated design. The current sensing is done by a LEM HAS 100-S hall effect current sensor. The output from this is fed into a potentiometer voltage divider to set the trip level. The MAX835 has an internal comparator with a reference voltage of 1.204V so if the voltage out of the pot is above that it will trip. The output is internally latched and can be reset by pressing the OCD reset switch or removing power. The output of this chip is active low so you can use it to control a transistor to shut off the signal to the gate drive when an event is happening.
I've created a very rough falstad simulation here if you would like the see how it works.
IGBT Selection
I'd like to take some time to go over the process of selecting a suitable IGBT switch for your DRSSTC, not just for a QCW design. The most reliable and quickest way to see if a certain IGBT will work for your application is to research what parameters people have previously tested the IGBT at. But if for whatever reason this is not possible to do, there is a rather straight forward but lengthy series of calculations that you can do to estimate the maximum switching frequency for given thermal values.
The formula below is for estimating the maximum switching frequency of an IGBT. It incorporates thermal losses and switching losses, rather than just using turn on delay, rise time, turn off delay, and fall time to calculate switching speed. I will try to describe how to extrapolate each value from a datasheet (BSM100GB60DLC), but if possible I recommend that you also read the original documentation by Microsemi.
fmax: The maximum frequency achievable before exceeding the given thermal values.
TJ: The junction temperature in °C.
TC: The temperature of the case of the IGBT in °C (heat spreader).
RθJC/ZthJC: The transient thermal impedance in K/W (or C/W depending on the datasheet. They are interchangeable).
Pcond: The conduction losses in W.
Eon/Eoff: The switching losses in mJ (or µJ depending on the datasheet).
RθJC can be found in the transient thermal impedance graph.
Since I am running a QCW that uses rf cycles lasting upwards of 20ms, I will be using a larger time value than a regular DRSSTC might. Since the duty cycle during the 20ms is 50%, I will use 10ms as my time. Looking at the graph we can see that my RθJC is about 0.075 K/W. (this whole part is a bit sketchy and it's debatable if it will produce completely accurate results. I've seen conflicting information on the method of calculating RθJC, this is just what worked for me)
Pcond is the sum of the collector current * the collector-emitter saturation voltage * D (our duty cycle from earlier) For IC I will be using the maximum value in the datasheet, of 200A. If necessary you could extend the graph to your desired current by continuing the slope. It is best to chose the values with the junction temperature closest to your maximum allowed temperature. Usually it will be 125C in the datasheet.
Switching losses can be found in the switching losses graph. Since this graph provides hard switching data, we can take only 20% to get a more accurate value. VGES is usually always 15V unless specifically specified. So we will need to do some manipulation of the values to get VGES at 24V.
Reading the graph we get Eon=2.1mJ, Eoff=4.9mJ @200A.
Now that we have all the values we can calculate fmax. We will use 80°C for TJ and 50°C for TC. These are generic values that you can change if you'd like.
It's important to note that there may be a somewhat large amount of discrepancy in Pdiss (the 400W value). It's best to expect around +-25% of variability with that and take the lowest value into consideration. So in my case 345kHz is the more conservative value.
All in all, actual real world experimentation is the only way to confidently know what switching speeds and IGBT can take. This calculation is only an estimate and sanity check.
Buck Converter
The buck converter may be the most simple part of the QCW. Mine uses a standard iron powder core, T300-2D, which seems to be the common choice amongst coilers. Given sufficient turns it will take a large amount of power without saturating. The switch used is a BSM150GB60DLC simply for the fact that it was cheap and the bricks that I already had on hand were unnecessarily large. As long as it's powerful enough, almost any IGBT can be used for this job since were only switching at 30kHz. Although we are hard switching so datasheet values will have to be respected for once. The beauty of using a dual IGBT module is that we can utilize the diode off the second IGBT as the diode in the buck converter. Of course it would be much better to just use the dual IGBT module as a synchronous buck converter, but I digress. The capacitor for the LC filter can be any power film cap. Usually DC link caps are a great choice.
Secondary and Primary Coil
Secondary: The secondary consists of 6" of 34AWG wire on a 3" ABS pipe. The carbon pigment content of this pipe is often said to conduct and burn easily in Tesla coil secondaries. Personally I have yet to encounter any problems relating to the ABS pipe. I even used it for my DRSSTC 2. That being said, It's preferable to use the white PVC pipe since some ABS manufactures may add more pigment than others. The secondary oscillates at around 390kHz unload (natural) and ~340kHz-360kHz loaded (upper pole) with a 250mm x 60mm toroid. For a QCW this is quite low but it will help to produce semi-fractal sparks.
Primary: Because of the necessity for a high coupling coefficient the primary will be helical and close to the secondary. And because of personal preference the primary is 14AWG insulated wire wound around a piece of PVC pipe. Tuning is still possible by removing a small amount of the upper turn.
More coming soon
References
[1] Jonathan Dodge, John Hess. IGBT Tutorial. 2002.
[2] Mads Barnkob. IGBT Selection for Tesla Coils or ZCS Inverters. 2015.
[3] Gao Guangyan. QCW Tesla Coil. 2014.